Pattern registration has always been a key part of successful photolithography. During the manufacture of integrated circuits many masks are used in succession and, in almost all cases, any given mask will need to be aligned relative to its predecessors with a degree of precision that is at least as well controlled as other pattern-related features such as line width.
A particular example of critical tolerances during pattern registration can be seen in the pattern format known as box-in-box. Referring now to FIG. 1a, we show an outer box 2 which has already been formed by photolithography. Inside 2 a smaller box 3 must be formed in a separate photolithography step. Both boxes have the shape of hollow squares and the tolerances involved are extremely tight. Typically, the long dimension 4 of the outer box would be about 20 microns while that of the inner box would be about 10 microns. Thus, the separation between boxes (for example in the region marked C) is typically about 3 microns.
In practice, to ensure perfect registration between masks, four box-in-box alignments are performed simultaneously, the boxes being located at the four corners of the image field. The outer box is formed from level 1 and the inner box from level 2.
What makes proper box-in-box registration particularly difficult is that the outer box can often be very difficult to measure because of deformation by the planarization process. This can be better understood by reference to FIG. 1b which is a cross-section of FIG. 1a through 1b--1b. The outer box is seen to be a metal filled trench 12 which has been overcoated with a second metal layer 15. The inner box is a photoresist pattern 18 which will be used to protect portions of layer 15 during etching. All that can be seen of the original trench 12 is a shallow step (such as 19) in the top surface of 15. To reliably detect the location of the inner photoresist pattern, an overlay measurement is used.
To better understand how difficulties arise during box-in-box overlay measurement it is instructive to review the two processes that currently precede the box-in-box process itself. Their common starting point is illustrated in FIG. 2 which shows trench 22 that has been formed in oxide layer 17 and then filled with layer 12 (which is typically tungsten although other metals could also have been used). Note that enough metal has been deposited to only partially fill 22. This is because the space is longer than twice the tungsten thickness and the tungsten film is highly conformal.
To remove excess metal everywhere except inside 22 and contact holes, one of two possible methods is used. FIG. 3a illustrates a common result when etch back (using a dry etch) is used. Layer 12 has been unevenly removed from inside 22 leaving behind residue 34 (which comprises tungsten and/or titanium nitride from the barrier layer) along with spacers 33 (that have been roughened as a result of grain growth) on the trench walls and with damage to the metal-oxide interface such as at 32, the latter damage resulting from etching during the over etch stage.
The alternative method for removing excess 12 is chemical mechanical polishing. The common result after using this method is illustrated in FIG. 3b. Although more of layer 12 is seen to have been left in 22, there is still a residue 35 of tungsten oxide (due to tungsten reaction with the slurry) left in the center of 22 and damage at the metal oxide interface, such as 23, can still readily occur.
FIGS. 4a and 4b show the appearance after deposition of the second metal layer (most commonly aluminum-copper, but other metals such as aluminum or aluminum-copper-silicon--with an Anti Reflection Coating of titanium nitride--could also have been used). For the etch back process (4a) as well as for chem. mech polishing (4b), the trench is seen to have been filled in an uneven and asymmetric manner because of the damage and poor filling of the previous step.
FIGS. 5a and b follow the two processes to the stage where the inner box (photoresist pattern 53) is now in place. For both processes, in order to correctly place 53, the overlay measuring tool traces are shown in FIGS. 6a and 6b. Sub-trace 61 for etch back is seen to be asymmetrical and to have an edge taper which is not sharp enough, and is therefore a potential source of error, while sub-trace 63 for chem. mech. polishing is seen to be unsuitable for the same reasons, usually even more serious in their manifestation.
In addition to the shortcomings in the prior art that lead to registration problems, as discussed above, the damage that occurs (see 23 and 32 above) can often lead to a break in the continuity of the outer box, as illustrated in FIG. 7 for outer box 72. The present invention seeks to remedy this problem as well as to greatly facilitate the registration problem.
Although we were unable to find any prior art that teaches the solution that will be described below, the following references were found to be of interest. Into (U.S. Pat. No. 4,938,600 July 1990) describes a method for displacement between layers wherein systematic errors associated with the measurement system are eliminated by measuring twice, with a 180.degree. rotation between measurements. Tanaka (U.S. Pat. No. 5,468,580 November 1995) shows how differences between successive overlay measurements can be minimized.
Nishimoto (U.S. Pat. No. 5,017,514 May 1991) uses a main vernier pattern formed at right angles to a subsidiary vernier pattern. Yim (U.S. Pat. No. 5,329,334 July 1994) describes a test reticle that includes a number of orthogonally arranged alignment marks of various shapes and sizes.